20 Countries Call for an International Body to Oversee the Most Powerful AI Models

Leaders of 20 countries have launched a joint initiative to strengthen international oversight of the most advanced artificial intelligence models. Among other things, they propose considering the creation of an international institution that could set safety standards and verify compliance by AI developers.

The initiative was presented on September 21 by Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre. Twenty-two political leaders from 20 countries, as well as European Commission President Ursula von der Leyen, joined the appeal.

Participants include Germany, Canada, Australia, Denmark, Spain, the Netherlands, Ireland, Estonia, Latvia, Moldova, Turkey, Singapore, Kazakhstan, Kenya, South Africa, the UAE, Bahrain, and other states.

The authors of the initiative demand that companies developing so-called frontier models—the most powerful and technologically advanced AI systems—create transparent safety protocols. These should include mandatory pre-deployment testing of models and independent risk assessment.

According to the proposal, independent experts must be granted sufficient access to the models to verify potential threats. Governments and regional organizations are urged to develop common standards and establish information sharing about serious AI incidents.

Another proposal concerns the creation of an international oversight mechanism. The countries called on UN members to explore the possibility of forming a special institution that could set standards, ensure independent verification, and convene states when AI systems reach a certain level of capability.

So far, this is specifically a political initiative: no decision has been made to create such a body, and its powers and structure have not yet been defined.

The joint appeal states that artificial intelligence must remain under human direction, supervision, and control and be used in accordance with international law.

The authors of the document explain the need for enhanced oversight by the rapid growth in capabilities of new systems. The appeal asserts that some current models have already demonstrated the ability to bypass safeguards during testing, exploit vulnerabilities, and gain unauthorized access to real systems.

Norwegian Prime Minister Jonas Gahr Støre said the potential benefits of AI remain enormous, but the technology’s growing capabilities create security risks that individual states cannot manage on their own. Discussion of the initiative will continue during the high-level week of the UN General Assembly in New York.

The document remains open for other states to join.

Based on materials from: President of Finland, Government of Norway